Files
impala/testdata/workloads/functional-query/queries/QueryTest
Tim Armstrong ce145ffee6 IMPALA-7600: bump mem_limit for test_kudu_scan_mem_usage
The estimate for memory consumption for this scan is 9 columns * 384kb
per column = 3.375mb. So if we set the mem_limit to 6.5mb, we should
still not get more than one scanner thread, but we can avoid hitting
out-of-memory.

The issue in the JIRA was queued row batches. With this change, and
num_scanner_threads=2, there should be max 12 row batches
(10 in the queue, 2 in the scanner threads about to be enqueued)
and based on the column stats I'd estimate that each row batch is
around 200kb, so this change should provide significantly more headroom.

Change-Id: I6d992cc076bc8678089f765bdffe92e877e9d229
Reviewed-on: http://gerrit.cloudera.org:8080/11513
Reviewed-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
Tested-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
2018-09-26 06:51:56 +00:00
..
2014-06-20 13:35:10 -07:00